Cem Iyigun

Comparison of missing value imputation methods in time series: the case of Turkish meteorological data

This study aims to compare several imputation methods to complete the missing values of spatio–temporal meteorological time series. To this end, six imputation methods are assessed with respect to various criteria including accuracy, robustness, precision, and efficiency for artificially created missing data in monthly total precipitation and mean temperature series obtained from the Turkish State Meteorological Service. Of these methods, simple arithmetic average, normal ratio (NR), and NR weighted with correlations comprise the simple ones, whereas multilayer perceptron type neural network and multiple imputation strategy adopted by Monte Carlo Markov Chain based on expectation–maximization (EM-MCMC) are computationally intensive ones. In addition, we propose a modification on the EM-MCMC method. Besides using a conventional accuracy measure based on squared errors, we also suggest the correlation dimension (CD) technique of nonlinear dynamic time series analysis which takes spatio–temporal dependencies into account for evaluating imputation performances. Depending on the detailed graphical and quantitative analysis, it can be said that although computational methods, particularly EM-MCMC method, are computationally inefficient, they seem favorable for imputation of meteorological time series with respect to different missingness periods considering both measures and both series studied. To conclude, using the EM-MCMC algorithm for imputing missing values before conducting any statistical analyses of meteorological data will definitely decrease the amount of uncertainty and give more robust results. Moreover, the CD measure can be suggested for the performance evaluation of missing data imputation particularly with computational methods since it gives more precise results in meteorological time series.

Clustering current climate regions of Turkey by using a multivariate statistical method

In this study, the hierarchical clustering technique, called Ward method, was applied for grouping common features of air temperature series, precipitation total and relative humidity series of 244 stations in Turkey. Results of clustering exhibited the impact of physical geographical features of Turkey, such as topography, orography, land–sea distribution and the high Anatolian peninsula on the geographical variability. Based on the monthly series of nine climatological observations recorded for the period of 1970–2010, 12 and 14 clusters of climate zones are determined. However, from the comparative analyses, it is decided that 14 clusters represent the climate of Turkey more realistically. These clusters are named as (1) Dry Summer Subtropical Semihumid Coastal Aegean Region; (2) Dry-Subhumid Mid-Western Anatolia Region; (3 and 4) Dry Summer Subtropical Humid Coastal Mediterranean region [(3) West coast Mediterranean and (4) Eastern Mediterranean sub-regions]; (5) Semihumid Eastern Marmara Transition Sub-region; (6) Dry Summer Subtropical Semihumid/Semiarid Continental Mediterranean region; (7) Semihumid Cold Continental Eastern Anatolia region; (8) Dry-subhumid/Semiarid Continental Central Anatolia Region; (9 and 10) Mid-latitude Humid Temperate Coastal Black Sea Region [(9) West Coast Black Sea and (10) East Coast Black Sea sub-regions]; (11) Semihumid Western Marmara Transition Sub-region; (12) Semihumid Continental Central to Eastern Anatolia Sub-region; (13) Rainy Summer Semihumid Cold Continental Northeastern Anatolia Sub-region; and (14) Semihumid Continental Mediterranean to Eastern Anatolia Transition Sub-region. We believe that this study can be considered as a reference for the other climate-related researches of Turkey, and can be useful for the detection of Turkish climate regions, which are obtained by a long-term time course dataset having many meteorological variables.

A generalized Weiszfeld method for the multi-facility location problem

An iterative method is proposed for the K facilities location problem. The problem is relaxed using probabilistic assignments, depending on the distances to the facilities. The probabilities, that decompose the problem into K single-facility location problems, are updated at each iteration together with the facility locations. The proposed method is a natural generalization of the Weiszfeld method to several facilities.

A genetic algorithm for the uncapacitated single allocation planar hub location problem

Given a set of n interacting points in a network, the hub location problem determines location of the hubs (transfer points) and assigns spokes (origin and destination points) to hubs so as to minimize the total transportation cost. In this study, we deal with the uncapacitated single allocation planar hub location problem (PHLP). In this problem, all flow between pairs of spokes goes through hubs, capacities of hubs are infinite, they can be located anywhere on the plane and are fully connected, and each spoke must be assigned to only one hub. We propose a mathematical formulation and a genetic algorithm (PHLGA) to solve PHLP in reasonable time. We test PHLGA on simulated and real life data sets. We compare our results with optimal solution and analyze results for special cases of PHLP for which the solution behavior can be predicted. Moreover, PHLGA results for the AP and CAB data set are compared with other heuristics.

Heuristics for a continuous multi-facility location problem with demand regions

We consider a continuous multi-facility location allocation problem where the demanding entities are regions in the plane instead of points. The problem can be stated as follows: given m (closed, convex) polygonal demand regions in the plane, find the locations of q facilities and allocate each region to exactly one facility so as to minimize a weighted sum of squares of the maximum Euclidean distances between the demand regions and the facilities they are assigned to.We propose mathematical programming formulations of the single and multiple facility versions of the problem considered. The single facility location problem is formulated as a second order cone programming (SOCP) problem, and hence is solvable in polynomial time. The multiple facility location problem is NP-hard in general and can be formulated as a mixed integer SOCP problem. This formulation is weak and does not even solve medium-size instances. To solve larger instances of the problem we propose three heuristics. When all the demand regions are rectangular regions with their sides parallel to the standard coordinate axes, a faster special heuristic is developed. We compare our heuristics in terms of both solution quality and computational time.

Probabilistic programming models for traffic incident management operations planning

This paper proposes mathematical programming models with probabilistic constraints in order to address incident response and resource allocation problems for the planning of traffic incident management operations. For the incident response planning, we use the concept of quality of service during a potential incident to give the decision-maker the flexibility to determine the optimal policy in response to various possible situations. An integer programming model with probabilistic constraints is also proposed to address the incident response problem with stochastic resource requirements at the sites of incidents. For the resource allocation planning, we introduce a mathematical model to determine the number of service vehicles allocated to each depot to meet the resource requirements of the incidents by taking into account the stochastic nature of the resource requirement and incident occurrence probabilities. A detailed case study for the incident resource allocation problem is included to demonstrate the use of proposed model in a real-world context. The paper concludes with a summary of results and recommendations for future research.

Restructuring forward step of MARS algorithm using a new knot selection procedure based on a mapping approach

In high dimensional data modeling, Multivariate Adaptive Regression Splines (MARS) is a popular nonparametric regression technique used to define the nonlinear relationship between a response variable and the predictors with the help of splines. MARS uses piecewise linear functions for local fit and apply an adaptive procedure to select the number and location of breaking points (called knots). The function estimation is basically generated via a two-stepwise procedure: forward selection and backward elimination. In the first step, a large number of local fits is obtained by selecting large number of knots via a lack-of-fit criteria; and in the latter one, the least contributing local fits or knots are removed. In conventional adaptive spline procedure, knots are selected from a set of all distinct data points that makes the forward selection procedure computationally expensive and leads to high local variance. To avoid this drawback, it is possible to restrict the knot points to a subset of data points. In this context, a new method is proposed for knot selection which bases on a mapping approach like self organizing maps. By this method, less but more representative data points are become eligible to be used as knots for function estimation in forward step of MARS. The proposed method is applied to many simulated and real datasets, and the results show that it proposes a time efficient forward step for the knot selection and model estimation without degrading the model accuracy and prediction performance.

The planar hub location problem: a probabilistic clustering approach

Given the demand between each origin-destination pair on a network, the planar hub location problem is to locate the multiple hubs anywhere on the plane and to assign the traffic to them so as to minimize the total travelling cost. The trips between any two points can be nonstop (no hubs used) or started by visiting any of the hubs. The travel cost between hubs is discounted with a factor. It is assumed that each point can be served by multiple hubs.We propose a probabilistic clustering method for the planar hub-location problem which is analogous to the method of Iyigun and Ben-Israel (in Operations Research Letters 38, 207–214, 2010; Computational Optmization and Applications, 2013) for the solution of the multi-facility location problem. The proposed method is an iterative probabilistic approach assuming that all trips can be taken with probabilities that depend on the travel costs based on the hub locations. Each hub location is the convex combination of all data points and other hubs. The probabilities are updated at each iteration together with the hub locations. Computations stop when the hub locations stop moving.Fermat-Weber problem and multi-facility location problem are the special cases of the proposed approach.

Determining the Climate Zones of Turkey by Center-Based Clustering Methods

There is a growing evidence that the climate change has already had significant impacts on the world’s physical, biological, and human systems, and it is expected that these impacts will become more severe in the near future. Alterations in the weather patterns and the existence of extreme events can be considered as important indicators of this change. The validity of this reality can be judged by analyzing climate data thoroughly. In this study, for determining the climate zones of Turkey, temperature measures obtained from the Turkish State Meteorological Service stations in the period 1950–2006 are examined by using two center-based clustering methods, namely k-means and fuzzy k-means. The clusters obtained from these methods are compared using objective criteria. They are also evaluated subjectively by the domain experts.

Efficient adaptive regression spline algorithms based on mapping approach with a case study on finance

Multivariate adaptive regression splines (MARS) has become a popular data mining (DM) tool due to its flexible model building strategy for high dimensional data. Compared to well-known others, it performs better in many areas such as finance, informatics, technology and science. Many studies have been conducted on improving its performance. For this purpose, an alternative backward stepwise algorithm is proposed through Conic-MARS (CMARS) method which uses a penalized residual sum of squares for MARS as a Tikhonov regularization problem. Additionally, by modifying the forward step of MARS via mapping approach, a time efficient procedure has been introduced by S-FMARS. Inspiring from the advantages of MARS, CMARS and S-FMARS, two hybrid methods are proposed in this study, aiming to produce time efficient DM tools without degrading their performances especially for large datasets. The resulting methods, called SMARS and SCMARS, are tested in terms of several performance criteria such as accuracy, complexity, stability and robustness via simulated and real life datasets. As a DM application, the hybrid methods are also applied to an important field of finance for predicting interest rates offered by a Turkish bank to its customers. The results show that the proposed hybrid methods, being the most time efficient with competing performances, can be considered as powerful choices particularly for large datasets.